“In February, the day after Sora came out, I warned that it didn’t really understand physics—and wouldn’t — in essay called Sora’s Surreal Physics:”
Our reading, the sentence that was checked:
Sora does not really understand physics and will not come to understand it.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Sora still appears to have trouble with physics, paragraph 1.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-sora-still-appears-to-have-trouble-c207f4-a1
We recorded 4 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-12-10. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
From the speaker
- Sora
- didn’t really understand physics—and wouldn’t
Our interpretation
- The forecast concerns Sora and subsequent versions of that model family, with no stated deadline.
- The plain reading of understanding physics is reliably generating physically consistent events across varied situations, including unfamiliar ones. It does not require flawless output.
- The historical premise and the unlimited prediction are distinct. Evidence of existing failures cannot establish the prediction.
Counts as supported if: Decisive evidence would have to establish an enduring inability of the relevant model family to achieve reliable physical consistency. Failures in finitely many versions provide provisional support only.
Counts as contradicted if: A later Sora version demonstrates reliable physical consistency and generalization across varied, independently evaluated situations.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Later research documents remaining coherence failures, while OpenAI reports improved physical accuracy. Neither finding settles an unlimited prediction about future understanding. No decisive event meeting the early-resolution rule was found.
Limitations
- Understanding has no speaker-specified test or threshold.
- The independent benchmark uses automated evaluators and tests interrupted state evolution rather than all of physics.
- Company claims of improvement are not independent validation.
Evidence (2 items)
Context only: Ziqi Ma, Mengzhan Liufu and Georgia Gkioxari / arXiv,
The coherence verifier captures these errors, with state-of-the-art closed-source models achieving only about a 60% success rate.
Why it matters here: The tested closed-source models include Sora 2 Pro. This reports substantial remaining coherence failures alongside successes; it does not establish permanent inability or provide a Sora-only rate in this sentence.
- Publisher
- Ziqi Ma, Mengzhan Liufu and Georgia Gkioxari / arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: OpenAI,
Our latest video generation model is more physically accurate, realistic, and more controllable than prior systems.
Why it matters here: OpenAI reports improved physical accuracy. This is material evidence against assuming no progress, but comparative improvement does not establish reliable understanding across unfamiliar situations.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
Independent, reproducible evaluations showing a later Sora version reliably generalizing physical behavior across varied unfamiliar situations would contradict the forecast. Further failures alone would leave its unlimited horizon open.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Sora still appears to have trouble with physics
In February, the day after Sora came out, I warned that it didn’t really understand physics—and wouldn’t — in essay called Sora’s Surreal Physics:
- Where
- paragraph 1
- Published
- 2024-12-10
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-sora-still-appears-to-have-trouble-c207f4-a1: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-sora-still-appears-to-have-trouble-c207f4-a1.